Integrating the Private Cloud with Existing Infrastructure, Part 1

The role of IT is changing. Whereas we were once the go-to guys and gals for all aspects of enterprise information automation, with a complete staff of in-house application experts, application expertise is now moving out to individual business units and departments. Simpler deployment and management interfaces reduce the technical skill level needed for application administration. But business depends much more on IT infrastructure to remain competitive and agile. Rapid deployment, predictable performance, reliability, and business continuity are common requirements that span all enterprise organizations. IT’s new role is to provide an information platform that has the capacity to meet response time requirements, a storage architecture to back up and restore applications at will, and end-to-end provisioning and management at the direction of application users.

The private cloud delivers all this through its user self-service portal. The cloud also adds the ability to accurately—and equitably—apportion costs among users and  to predict future resource requirements. But few of us get to start with a clean sheet of paper when planning any IT paradigm shift. We must accommodate existing infrastructure, applications, and processes. Thus, although private clouds are becoming increasingly affordable and easy to deploy—you can now buy “cloud-in-a-box” solutions—the process of integrating a private cloud with existing enterprise infrastructure requires careful planning. You should focus on three areas: service orchestration, self-service interfaces, and monitoring capabilities. In this post, I'll talk about the first area; I'll discuss the others in a follow-up post next week.

Orchestrating Change

Traditional IT provisioning of new application environments and infrastructure resources entails many planning, collaboration, approval, and technical steps. End users relay their requirements to a project manager or app developer, who then requests the necessary software, server, networking, and storage resources.  A system administrator builds out those resources—typically in a virtualized data center—and an operations team moves the resulting application package into production. The key objective of private cloud integration is to transition those existing processes, collectively termed orchestration, to the automated provisioning process that puts end users in charge of application-deployment details.

Whereas traditional IT orchestration is a manual process, the private (and public) cloud automates orchestration to let IT deliver entire business services as a consistent package. The first step, then, toward private cloud orchestration is to identify the most commonly used business servers users require, and focus on making them into private cloud-orchestrated services. For your initial private cloud offerings, you might consider generic services, such as email, document management, or web content delivery. (Take our current instant poll to vote on the application that you'd pick first.) Many cloud platforms already support a wide range of fully orchestrated workflows for such generic applications, and most businesses have ongoing needs to deploy them. Generic app delivery through a private cloud lets you, and your users, see how a self-service portal can make IT more responsive and lets users deploy applications according to their own availability, rather than that of the IT staff.

After you gain experience with orchestration with a few generic apps, you can move to line-of-business apps such as accounting, CRM, or ERP. You might find that your existing software suppliers already offer private cloud versions of their application stacks. Or you might use private cloud deployment as an opportunity to migrate a major application to a new software platform. Either approach will give your users more confidence in private cloud processes and results, while moving a major portion of your business to your new internal cloud.

As you catalog existing applications and begin moving them to a service (rather than a component) model, plan the orchestration process to minimize the number of human interventions. You can do this by moving all approval steps to the front of your orchestration workflow, so that final provisioning is a fully automated process.

Next week, I'll take a look at private cloud self-service and monitoring.